Fiziksel insan robot etkileşiminin denetleyici tasarımı için çok kriterli bir optimizasyon çerçevesi

In the near future, humans and robots are expected to perform collaborative tasks involving physical interaction in various environments, such as homes, hospitals, and factories. Robots are good at precision, strength, and repetition, while humans are better at adaptation and cognitive tasks. Physical human-robot interaction (pHRI) takes advantage of both robots and humans to improve speed, flexibility, and ergonomics of complex tasks. pHRI requires design of controllers to achieve safe and transparent operations which is challenging mainly due to the contradicting nature of these objectives. Knowing that attaining perfect transparency is practically unachievable, controllers allowing better compromise between these objectives are desirable. In the first part of this dissertation, we propose a new controller, fractional order admittance controller, for pHRI. The stability and transparency analyses of the new control system are performed computationally with human-in-the-loop. Impedance matching is proposed to map fractional order control parameters to integer order ones to enable comparisons, and the stability robustness of the system is studied analytically. Furthermore, the interaction performance is investigated experimentally through two human subject studies involving continuous contact with linear and nonlinear viscoelastic environments. The results indicate that the fractional order admittance controller can be more robust and transparent than the integer order admittance controller and the use of fractional order term can reduce the human effort during tasks involving contact interactions with environment. In the second part of this dissertation, we propose a multi-criteria optimization framework for interaction controller design, which jointly optimizes the stability robustness and transparency of a closed-loop pHRI system for a given interaction controller structure. In particular, we propose a Pareto optimization framework that allows the designer to make informed decisions by studying the tradeoff between stability robustness and transparency thoroughly. The proposed framework involves a search over the discretized controller parameter space to compute the performance metrics. Using multi-criteria optimization, Pareto front curve depicting the tradeoff between stability robustness and transparency is obtained. Studying this tradeoff, an informed decision can be made to select the set of controller parameters that yield maximum attainable transparency and stability robustness. To demonstrate the practical use of the proposed design approach, integer and fractional order admittance controllers are studied as a case study and compared both analytically and experimentally. Experimental comparisons are performed for a pHRI task that involves contact interactions with an environment displaying nonlinear stiffness. The results validate the proposed design framework, and show that the achievable transparency under fractional order admittance controller is higher than that of integer order one, when both controllers are designed to ensure the same level of stability robustness.